The Client
MacadamIA is an AI-powered energy advisor for the Spanish electricity market. Founded by Luis Manzano Herrero, the platform helps consumers find better electricity tariffs and switch providers — entirely online, without phone calls or paperwork.
Spain's liberalized energy market has over +20 providers, each with multiple tariff types across three consumption periods: Punta (peak), Llano (flat), and Valle (off-peak). For most consumers, this means navigating confusing rate tables and dealing with aggressive telemarketing. MacadamIA was built to change that — and Luis needed a small, senior team that could move fast and own the entire stack, from AI to infrastructure.
The Challenge
The core problem was deceptively complex: take a PDF electricity invoice — different format for every provider, no standardization — and turn it into a personalized tariff recommendation. The LLM had to work reliably with real, messy documents in production. Tariff calculations required applying Spanish taxes (5.18% + 21% VAT) across three time-of-use periods, with real pricing from 20 providers. And sensitive financial data (IBAN, national ID) needed GDPR-compliant encryption.
On the business side, Spanish consumers distrust energy companies after years of aggressive sales tactics. MacadamIA needed to build trust through transparency and scale lead capture without a call center — a fully digital model from first interaction to provider switch.
"Truly understood our business needs. Delivered the exact AI-powered MVP required, on time, with real empathy throughout the process."
The Solution
Consumers used to spend hours comparing confusing tariff tables, only to end up on the phone with aggressive salespeople. We built a system where a quick chat with an AI assistant, a bill upload, and a few clicks replace the entire process — the kind of AI system that works with real data, real users, and real edge cases from day one.
AI-Powered Invoice Analysis
We built Maca — a conversational AI assistant powered by OpenAI GPT with streaming responses via Vercel AI SDK. Users upload their electricity bill (PDF or photo) directly in chat, and the LLM extracts what matters: account holder, consumption per period in kWh, contracted power in kW, the supply point identifier (CUPS), and current charges.
What makes this work in production — not just in demos — is how it handles edge cases. Gas invoices? Maca recognizes them and redirects. Business tariffs on 3.0TD contracts? Flagged and handled differently. Indexed tariffs with volatile pricing? Maca educates the user on the trade-offs before recommending. The entire process is conversational — no forms, no friction.
"I recommended it to my sister, my neighbor, and the women in my yoga group. They switched my tariff six months ago and I'm paying significantly less. No lock-in, no fine print, no surprises."
View original (Spanish)
"Se lo recomendé a mi hermana, a mi vecina y a las del grupo de yoga. A mí me cambiaron la tarifa hace ya medio año y estoy pagando bastante menos. Sin permanencia, sin letra pequeña, sin sorpresas."
Tariff Calculation Engine
Behind the conversational interface sits a calculation engine with real pricing data from +20 Spanish energy providers. For each user's consumption profile, it computes monthly and annual costs — applying the correct electricity tax and VAT — across every available tariff. Results are ranked by user preference: fixed price for predictability, time-of-use for optimization, or indexed rates for those comfortable with market volatility. Each option shows the estimated savings compared to the user's current tariff, displayed as interactive cards users can compare side by side. This isn't a lookup table — it's real-time, regulation-compliant math across +20 providers.
Digital Lead Capture Flow
When a user selects their preferred tariff, the switching process stays entirely digital. The flow moves from chat to a secure completion form — accessed via unique, time-limited tokens — where users provide the remaining details needed to initiate the switch. For users who don't complete immediately, automated reminder emails trigger with escalating urgency. When a lead completes the full flow, the admin team is notified instantly. Zero phone calls, zero paperwork — from first chat message to provider switch request, everything happens online. No call center needed.
"The team quickly grasped the business idea and collaborated closely with the client. Their empathy set them apart."
Infrastructure & Operations
MacadamIA runs on Next.js 16 with React 19 and TypeScript, backed by PostgreSQL with Drizzle ORM. Containerized with Docker, served through Nginx with security headers, and deployed via GitHub Actions CI/CD. Sensitive data is encrypted with AES-256-CBC (GDPR compliant), errors tracked through Sentry, and analytics handled by PostHog.
We also built a separate admin backoffice — a full dashboard where MacadamIA's team manages leads, updates tariff pricing across all +20 providers, monitors conversion analytics, and publishes blog content. No developer needed for day-to-day operations.
Want results like these?
MacadamIA went from idea to production AI in 12 weeks. Start with a Feasibility Report — a one-week deep dive into your use case with architecture recommendations and a working prototype.
The Results
MacadamIA launched with a 30%+ switch rate — more than 3 in 10 users who analyze their bill initiate a provider switch. In an industry where conversion typically requires phone calls and aggressive sales, this is achieved through a purely digital, trust-based flow.
80%+ of users interact with Maca for personalized advice. This isn't a chatbot that gets ignored — it's the core product experience. Users trust it to analyze their real invoices and recommend real savings.
The switching process went from days to minutes. What previously required phone calls, mailed paperwork, and weeks of waiting is now a conversation and a form — fully digitalized, end to end.
The platform compares tariffs from +20 providers in real time, with accurate tax calculations for the Spanish market. Users report savings of up to €420 per year.
All of this was built and shipped in 12 weeks by a dedicated senior engineering team — no ML specialists, no project managers, no overhead. Just direct collaboration between founders and engineers.
"With two kids at home, the electricity bill really adds up. We were paying around €130/month and it found us a tariff that brought it down to about €95. It also explained that not all tariffs work for everyone — that's something you appreciate."
View original (Spanish)
"Con dos niños en casa la factura de la luz se nota mucho. Estábamos pagando unos 130€ al mes y nos encontró una tarifa que nos la ha bajado a unos 95€. También me explicó que no todas las tarifas sirven para todos, eso se agradece."